The New Sudoku Players' Forum

[ronk]

I have imagined to add 3400 grids to the gordon file It's a little more than 10%

Suggest you post about 50 of them here to confirm your results.

[ravel]

I begin to generate 17 clues but my generator is slow: 20 by hour

wow, dukuso himself would have been happy with this rate. Are you sure?

[Papy]

Here are the first grids
They are nit verified(I wait to have more to study them)

but when I generate them I solve them and verify the uhnique solution
Some can be in the Gordon files they are not sorted

Code: Select all

917356482352784619486219735675932148238541976194867523841625397763498251529173864
91___________8_6__4___________9_2_4__38_____________2_______3_77__4________1__8__
915367482362584719487219536756932148238641957194875623841726395573498261629153874
91___________8_7__4___________9_2_4__38_____________2_______3_55__4________1__8__
917356482352784619486219735765932148238541976194867523841625397573498261629173854
91___________8_6__4___________9_2_4__38_____________2_______3_75__4________1__8__
915376482372584619486219537567932148238741956194865723841627395653498271729153864
91___________8_6__4___________9_2_4__38_____________2_______3_56__4________1__8__
915376482372584619486219537567932148238741956194865723841627395753498261629153874
91___________8_6__4___________9_2_4__38_____________2_______3_57__4________1__8__
916357482352684719487219635675932148238541967194876523841725396563498271729163854
91___________8_7__4___________9_2_4__38_____________2_______3_65__4________1__8__
917365482362784519485219736756932148238641975194857623841526397573498261629173854
91___________8_5__4___________9_2_4__38_____________2_______3_75__4________1__8__
916375482372684519485219637567932148238741965194856723841527396653498271729163854
91___________8_5__4___________9_2_4__38_____________2_______3_66__4________1__8__
916375482372684519485219637657932148238741965194856723841527396763498251529163874
91___________8_5__4___________9_2_4__38_____________2_______3_67__4________1__8__
918267345675438291432159867827341659369875412541692738154786923286913574793524186
9_8__73__6____________5_____2__4__5_3__8___________7___54____2____9_3____________
968127345175438296432659871827346159319875462546291738654782913781963524293514687
9_8__73__1____________5_____2__4__5_3__8___________7___54____1____9_3____________
928167345175438692436259871867342159319875426542691738254786913781923564693514287
9_8__73__1____________5_____6__4__5_3__8___________7___54____1____9_3____________
968217345175438296432659817827346159319875462546192738654781923281963574793524681
9_8__73__1____________5_____2__4__5_3__8___________7___54____2____9_3____________


Plese
'ne tirez pas sur le pianiste'
In fre'nch in the text!
(If some (all) grids are bad please dont' kill me!!!

Papy

[ronk]

Only puzzles 1, 3, 4, 10, and 11 are mutually non-equivalent. In that order, they are:

Code: Select all

91...........8.6..4...........9.2.4..38.............2.......3.77..4........1..8..
91...........8.6..4...........9.2.4..38.............2.......3.75..4........1..8..
91...........8.6..4...........9.2.4..38.............2.......3.56..4........1..8..
9.8..73..6............5.....2..4..5.3..8...........7...54....2....9.3............
9.8..73..1............5.....2..4..5.3..8...........7...54....1....9.3............

I could not find any of them in Gordon's collection of 36628 puzzles with 17 clues ... but then I've never done that step before, so I could easily be wrong.

[ravel]

Papy, when you do what gsf suggested above - download his solver, write your puzzles to a file papynew17.dat and (in a cmd shell in the directory with the solver) call
sudoku -qFN -f%#0c papynew17.dat > papynew17.can
you will find that all but 5 of the puzzles in papynew17.can are equal.
They also can be found in Gordons (canonicalized) list.

[Papy]

I tell you that the my grids have not been controled in the Gordon file!
I just generate valid 17 clues and is't normal taht some of my solutions are in the goirdon but now I control at the generation is the grid is good.

I don't find the link to the code or algo to c14n Perhaps it will be faster to c14n my solution before solve them.
I write a very stupid but fast solver but sometimes it's slow.
I don't understand when.
The speed is slow only on some ,not all, 17 clues.
But I think I'm on the good way
if 5 are good it's then beginning

Papy

[Red Ed]

But I think I'm on the good way
if 5 are good it's then beginning

You're right that if you get a few 17s then it's a good beginning, since you can often use those to generate new puzzles.

Say that two 17s are "related" (just for the purposes of this posting) iff you can get from one to the other by isomorphism and up to 3 edits, where an edit is a cell-write operation grid[r][c] = v (v=0 means "empty"). This relation defines a graph of 36628 vertices (known 17s) partitioned into, it happens, 13517 components. The largest component has size 8980, encompassing nearly a quarter of all known 17s.

If you get lucky and find a 17 in this component then the remaining 8979 17s are yours for the taking simply by trying up to 3 edits for each new one found. It'll take a while, obviously, and I'm sure Gordon did something far cleverer than that, but it's guaranteed to give you loads of 17s eventually.

[ronk]

I could not find any of [the five non-equivalents] in Gordon's collection of 36628 puzzles with 17 clues ... but then I've never done that step before, so I could easily be wrong.

you will find that all but 5 of the puzzles in papynew17.can are equal.
They also can be found in Gordons (canonicalized) list.

if 5 are good it's then beginning

Sorry, none of your 13 puzzles are "good". I found my error and, as ravel says, each of the 5 mutually non-equivalents is already in Gordon's (canonicalized) list. In order of their appearance in Gordon's list, the isomorphs are puzzles 36574, 36575, 36578, 36579 and 36580.

[ravel]

The speed is slow only on some ,not all, 17 clues.

Do you mean the solving speed ? Then i also can recommend gsf's solver.

But I think I'm on the good way
if 5 are good it's then beginning

Think there are not much people, who generated a 17 clue. So if your search is independant enough from the known ones, you are on a good way and even should have chances to find a new one.

[Papy]

News grids.
I take a filter with the gordon solutions but not perfect
(I have not the time to write a C14n)
I work on the generation

Code: Select all

9_1_________7__3________4__6___19_8__3____2_7__________7_2____________96___4_____
9_1_________7__3________4__6___19_8__3____2_7__________7_2____________96___4_____
9_1_________7__3________2__6___19_8__3____4_7__________7_2____________96___4_____
9_1_________3___6_______________51_263_8_______________8_61____4______5_____2_9__
9_1_________3___6_______________51_263_7_______________7_61____4______5_____2_9__
9_1__________6_2_____38_____8____65____4__3_____9__________4_1923________________
9_1__________6_2_____38_____8____65____7__3_____9__________7_1923________________
9__87__________4_1___9__2__32__________6___8______1______5__3___14________7______
9__87__________4_1___9__2__32__________7___8______1______6__3___14________7______
9__81__________5_2___9__3__43__________1___8______2______6__4___25________1______
9__8_1_7_4_____3___________6__93___________81____4_____715___________9___8_______
9__8_1_7_4_____3___________2__93___________81____4_____716___________9___8_______
9__8__23__7__1__________9__3_2__4_8_____6____8_________6______1___2________3_____
9__8__23__5__1__________9__3_2__7_8_____6____8_________6______1___2________3_____
9__7___4__2__3______________53___2___8_4_6______1_____4______6_____8_5__1________
9__7___4__2__3______________53___2___8_4_6______1_____4______9_____8_5__1________
9__7_1____3______8_______4__4__8__3____3__1__2___________52____1_____9______4____
9__6_1____3______8_______4__4__8__3____3__1__2___________72____1_____9______4____
9__5__3_____4_1____________59_7_______4_____1____3_2__63__2___________5________4_
9__5___________7_________8____6__4_9_82_________1__5__43__28_________6______3____
9__5___________7_________8____1__4_9_82_________6__5__43__28_________6______3____
6__4__7___1______2___________2___3_____1___5__8_______1_____94_____27__85________
9__4__7___1______2___________2___6_____1___5__8_______1_____94_____27__85________
9__4__7___1______2___________2___3_____1___5__8_______1_____64_____27__85________
8__4__6___1______2___________2___3_____1___5__7_______1_____84_____26__75________
9__4__3______8___5____________9__74__51_________2_____7___53____3__1___________2_
9__4___3__8__7_5____1______2__9_3____7____8_1__________5__8____4______9__________
93__18__________2_____3__________1_82__5_____6___________2_5_7__1____9_____6_____


Papy

[Papy]

Sorry I don't understand your demonstartion...

My method will give a limlited set of grid
I don't know how manyt % will there in the gordon collection
I don't use isomorphs to generate...

I will be intrested by my bad solutions and the Gordon puzzle.
Using my way the twins must be twin So if not I'm ionterrested
to understand why not.

Sgf:
If you have tipos to eliminate bad solutions DURING the elaboration I'm interrested!
Thanks

Papy

[ravel]

19 non isomorphic 17-clues, all in Gordon's list, half of them at the very end (last 500?).
You should check this yourself, it can be done with free programs without big effort.

[Papy]

I wait to have a big file to test wiht the gordon because I want to write a c14n and I cannot modify the generator
An controlling them a true time willl slow my generator

If the grids are in the gordon files no mater
Gordon make a collection of grids and me I try to generate
Suer if I generate new grid i will be happy but generating 2000 or 3000
17 clues id not so bad!

My first genratot makes invalid grid
This onbe make goods
It'a a progress
Papy

[gsf]

Sgf:
If you have tipos to eliminate bad solutions DURING the elaboration I'm interrested!

is a "bad solution" a puzzle already in gordon's list? or simply an invalid sudoku?

[Papy]

Hi,
before to dispose the clues I make a valid grid

Starting from zero

I fill the row one by one directly with a valid row (I compute all)
When i put the second I control if the celles of the second row is compatible with the cells of the first

123 456 789
987 513 426

is bad because th 5 is two times in the box2
also because youy have 82(C2) and 28(C8) and it's impossible(multiply solutions)
I look from tips for two row or a blocks.
Do you use this kind od detections

For C14n : looking at the G file it doesn't seem that the grid are from the same familly
C14n is just a calcul on the grids not a specific managements of the cell?

Papy